Post on 19-Mar-2020
The Future of Swiss Railway Dispatching. Deep Learning and Simulation on DGX-1.
Adrian Egli & Erik Nygren
Research and Innovation Platform SBB AG, Switzerland
Swiss Federal Railways. Complex dynamics in the heart of Europe.
Basic train dispatching. Reordering of trains.
Basic train dispatching. Rerouting of trains.
Train runs. A simple chain of dispatching decisions.
Interacting trains. The source of railway complexity.
Train runs. A path in a decision tree.
Most dense mixed train network in the world. Exponential growth of complexity.
1 2 80 4
30 8 Mio. 900
>80
?
~10 80 Mio.
Sensitive dynamical system. Finding the needle in the haystack.
Increasing future mobility needs. Destabilizing effects of traffic density.
Future Today
Maintaining robust traffic flow. Increased man- and computational power.
Future
+ +
Maintaining robust traffic flow. Infrastructure enhancements.
Future
+
Future projections. Inevitable challenges.
Time
Performance
Cost
Quality
Traffic density
Overcoming future challenges. Making the railway network antifragile.
Antifragility
Antifragility. Improvement through failure.
Antifragility. Improvement through failure.
How to fail in a safe way. Extending the railway network beyond reality.
Simulation
Validation
Dispatcher
Swiss Railway Digital Twin. Infinite possibilities.
Reinforcement learning. Mastering complex games.
Game
Agent
Reinforcement learning. Playing the dispatcher game.
Railway simulation
Agent
Super human performance. Learning from 65 million years of experience.
65 Mio. years
High performance simulations. The power of parallel computations.
python
PyCUDA
Digital Twin. Moving beyond the physical boundaries.
Digital Twin. Moving beyond the physical boundaries
High performance simulations. State of the art.
Reinforcement
Learning
13.9 sec.
Business
Rules
2.8 sec.
Physics
Simulation
0.3 sec.
Swiss railway
network
17 sec.
1
31K 15K
13K 800
Learning from 65 million years of experience. Time as a limiting factor.
65M years
experience
12K years
training 17s
1
= x
1
Limited time resources. Scaling with innovative ideas.
GPU
Agent
Agent
Agent
Railway simulation. Learning on subregions.
Railway simulation. Reinforcement agents view.
High performance computing. Parallel training on alternative worlds.
Diversity, curiosity, passion and team work. The evolution of a digital twin.
Deep learning and simulation. The (r)evolution of the Swiss Federal Railways.
Reality Digital Trial & Error
Erik Nygren
erik.nygren@sbb.ch
AI Researcher
Adrian Egli
adrian.egli@sbb.ch
HPC Expert
Dirk Abels
dirk.abels@sbb.ch
Head of Research Lab
Research Team. Pushing railway to the next level.